Public Response and Emotional Evolution in Extreme Precipitation Events Based on Social Media Big Data: A Case Study of “7·31” Heavy Rain in Beijing in 2023

Based on Weibo big data, BERTopic, and dual-channel sentiment analysis model, a dynamic analysis framework of public perception and emotion evolution is constructed from the perspective of disaster chain and public response. The results show that (1) Due to the trust of information sources by the public, the efficiency of early warning information reaching the public and attracting attention is relatively high. Social media activity on related topics peaked several times in response to reports of major hazards, such as railway suspensions, passengers trapped in trains, and severe flooding in Miyun District, Beijing. (2) The evolution of topics of public attention strongly corresponds to the disaster process: From early warning and emergency risk avoidance, gradually move to disaster report and rescue coordination, and finally focus on the criticism of infrastructure vulnerability. (3) The emotional response presents phased characteristics. At the initial stage of the rainstorm red warning issued by the meteorological department, anxiety is dominant; after the release of rescue information, emotion rises briefly, and reflection and attribution tendencies generally appear after the disaster. (4) Elderly populations and those in remote areas exhibit characteristics of vulnerability, information isolation, and high dependency in response to disasters.